Revert URL subtraction

It appears we merged the URL subtraction a bit too soon, and we need
to rework the approach. I know have ~20 hours into it trying to solve
the issues. The conclusion is that it is not ready to ship and we need
a new approach.

To ensure it does not go out before its ready, revert the original PR
and the attempts to solve the issues with it so we can start fresh.

References

docs/api.rst

@@ -997,21 +997,6 @@ The path is encoded if needed.
          >>> base.join(URL('//python.org/page.html'))
          URL('http://python.org/page.html')
 
-The subtraction (``-``) operator creates a new URL with
-a relative *path* to the target URL from the given base URL.
-*scheme*, *user*, *password*, *host*, *port*, *query* and *fragment* are removed.
-
-.. method:: URL.__sub__(url)
-
-   Returns a new URL with a relative *path* between two other URL objects.
-
-   .. doctest::
-
-      >>> target = URL('http://example.com/path/index.html')
-      >>> base = URL('http://example.com/')
-      >>> target - base
-      URL('path/index.html')
-
 Human readable representation
 -----------------------------
 

tests/test_url.py

@@ -61,89 +61,6 @@ def test_str():
     assert str(url) == "http://example.com:8888/path/to?a=1&b=2"
 
 
-@pytest.mark.parametrize(
-    ("target", "base", "expected"),
-    [
-        ("http://example.com/path/to", "http://example.com/", "path/to"),
-        ("http://example.com/path/to", "http://example.com/spam", "path/to"),
-        ("http://example.com/this/is/a/test", "http://example.com/this/", "is/a/test"),
-        (
-            "http://example.com/this/./is/a/test",
-            "http://example.com/this/",
-            "is/a/test",
-        ),
-        ("http://example.com/this/is/../a//test", "http://example.com/this/", "a/test"),
-        ("http://example.com/path/to", "http://example.com/spam/", "../path/to"),
-        ("http://example.com/path", "http://example.com/path/to/", ".."),
-        ("http://example.com/path", "http://example.com/other/../path/to/", ".."),
-        ("http://example.com/", "http://example.com/", "."),
-        ("http://example.com", "http://example.com", "."),
-        ("http://example.com/", "http://example.com", "."),
-        ("http://example.com", "http://example.com/", "."),
-        ("//example.com", "//example.com", "."),
-        ("/path/to", "/spam/", "../path/to"),
-        ("path/to", "spam/", "../path/to"),
-        ("path/../to", "path/", "../to"),
-        ("path/..", ".", "path/.."),
-        ("path/../replace/me", "path/../replace", "replace/me"),
-        ("path/../replace/me", "path/../replace/", "me"),
-        ("path/to", "spam", "path/to"),
-        ("..", ".", ".."),
-        (".", "..", "."),
-    ],
-)
-def test_sub(target: str, base: str, expected: str):
-    expected_url = URL(expected)
-    result_url = URL(target) - URL(base)
-    assert result_url == expected_url
-
-
-@pytest.mark.xfail(reason="Empty segments are not preserved")
-@pytest.mark.parametrize(
-    ("target", "base", "expected"),
-    [
-        (
-            "http://example.com/////path/////to",
-            "http://example.com/////spam",
-            "path/////to",
-        ),
-        (
-            "http://example.com////path/////to",
-            "http://example.com/////spam",
-            "..//path/////to",
-        ),
-    ],
-)
-def test_sub_empty_segments(target: str, base: str, expected: str):
-    expected_url = URL(expected)
-    result_url = URL(target) - URL(base)
-    assert result_url == expected_url
-
-
-def test_sub_with_different_schemes():
-    expected_error_msg = "Both URLs should have the same scheme"
-    with pytest.raises(ValueError, match=expected_error_msg):
-        URL("http://example.com/") - URL("https://example.com/")
-
-
-def test_sub_with_different_netlocs():
-    expected_error_msg = "Both URLs should have the same netloc"
-    with pytest.raises(ValueError, match=expected_error_msg):
-        URL("https://spam.com/") - URL("https://ham.com/")
-
-
-def test_sub_with_different_anchors():
-    expected_error_msg = "'path/to' and '/path' have different anchors"
-    with pytest.raises(ValueError, match=expected_error_msg):
-        URL("path/to") - URL("/path/from")
-
-
-def test_sub_with_two_dots_in_base():
-    expected_error_msg = "'..' segment in '/path/..' cannot be walked"
-    with pytest.raises(ValueError, match=expected_error_msg):
-        URL("path/to") - URL("/path/../from")
-
-
 def test_repr():
     url = URL("http://example.com")
     assert "URL('http://example.com')" == repr(url)

tests/test_url_benchmarks.py

@@ -589,20 +589,6 @@ def _run() -> None:
             url.query
 
 
-def test_url_subtract(benchmark: BenchmarkFixture) -> None:
-    @benchmark
-    def _run() -> None:
-        for _ in range(100):
-            URL_WITH_LONGER_PATH - URL_WITH_PATH
-
-
-def test_url_subtract_long_urls(benchmark: BenchmarkFixture) -> None:
-    @benchmark
-    def _run() -> None:
-        for _ in range(100):
-            URL_VERY_LONG_PATH - URL_LONG_PATH
-
-
 def test_url_host_port_subcomponent(benchmark: BenchmarkFixture) -> None:
     cache_non_default = URL_WITH_NOT_DEFAULT_PORT._cache
     cache = BASE_URL._cache

yarl/_path.py

@@ -2,8 +2,6 @@
 
 from collections.abc import Sequence
 from contextlib import suppress
-from itertools import chain
-from pathlib import PurePosixPath
 
 
 def normalize_path_segments(segments: Sequence[str]) -> list[str]:
@@ -41,31 +39,3 @@ def normalize_path(path: str) -> str:
 
     segments = path.split("/")
     return prefix + "/".join(normalize_path_segments(segments))
-
-
-def calculate_relative_path(target: str, base: str) -> str:
-    """Return the relative path between two other paths.
-
-    If the operation is not possible, raise ValueError.
-    """
-
-    target = target or "/"
-    base = base or "/"
-
-    target_path = PurePosixPath(target)
-    base_path = PurePosixPath(base)
-
-    if base[-1] != "/":
-        base_path = base_path.parent
-
-    for step, path in enumerate(chain((base_path,), base_path.parents)):
-        if path == target_path or path in target_path.parents:
-            break
-        elif path.name == "..":
-            raise ValueError(f"'..' segment in {str(base_path)!r} cannot be walked")
-    else:
-        raise ValueError(
-            f"{str(target_path)!r} and {str(base_path)!r} have different anchors"
-        )
-    offset = len(path.parts)
-    return str(PurePosixPath(*("..",) * step, *target_path.parts[offset:]))

yarl/_url.py

@@ -12,7 +12,7 @@
 from propcache.api import under_cached_property as cached_property
 
 from ._parse import USES_AUTHORITY, make_netloc, split_netloc, split_url, unsplit_result
-from ._path import calculate_relative_path, normalize_path, normalize_path_segments
+from ._path import normalize_path, normalize_path_segments
 from ._query import (
     Query,
     QueryVariable,
@@ -477,34 +477,6 @@ def __truediv__(self, name: str) -> "URL":
             return NotImplemented
         return self._make_child((str(name),))
 
-    def __sub__(self, other: object) -> "URL":
-        """Return a new URL with a relative path between two other URL objects.
-
-        Note that both URLs must have the same scheme and netloc.
-        The new relative URL has only path:
-        scheme, user, password, host, port, query and fragment are removed.
-
-        Example:
-        >>> target = URL("http://example.com/path/index.html")
-        >>> base = URL("http://example.com/")
-        >>> target - base
-        URL('path/index.html')
-        """
-
-        if type(other) is not URL:
-            return NotImplemented
-
-        target_scheme, target_netloc, target_path, _, _ = self._val
-        base_scheme, base_netloc, base_path, _, _ = other._val
-
-        if target_scheme != base_scheme:
-            raise ValueError("Both URLs should have the same scheme")
-        if target_netloc != base_netloc:
-            raise ValueError("Both URLs should have the same netloc")
-
-        path = calculate_relative_path(target_path, base_path)
-        return self._from_tup(("", "", path, "", ""))
-
     def __mod__(self, query: Query) -> "URL":
         return self.update_query(query)